or Connect
AVS › AVS Forum › Display Devices › Display Calibration › HCFR - Open source projector and display calibration software
New Posts  All Forums:Forum Nav:

HCFR - Open source projector and display calibration software - Page 89

post #2641 of 3446
Thread Starter 
Quote:
Originally Posted by MonarchX View Post

Could you link me to a guide on 1D LUT vs. 3D LUT? It is awfully confusing to me - I need simple terms. I thought that the whole idea of D65 REC. 709 standard was that it was a standard to make all displays look the same so that you can view the image in its original form. I can correct all of my color spaces with software without having to use 3D LUT because everything I have done now was through hardware. What I can't figure out is whether there is a difference between theoretical 100% perfectly accurate calibration using 1D LUT and theoretical 100% perfectly accurate calibration using 3D LUT? A standard is a standard.. or not?

People often mix up the terms 1D and 3D. A gray scale calibration is called 1D because it corrects only one axis of your color space. The CMS controls on your display (hue, sat, lightness) are actually 3D but they contain only 6 control points. These 6 control points usually aren't enough to remove all errors with respect to rec.709 or whatever color space you are aiming for. The madVR 3DLUT(software) and eeColorbox 3DLUT (hardware) contain many thousands of control points that can be used to correct color. The basic process work flow is:


Characterize your display errors (profile) -> calculate what adjustment need to be made to the source to remove errors at the display->Load those adjustments into your 3DLUT device.
post #2642 of 3446
Thanks!

Games and movies now look so damn amazing on my HDTV now!
post #2643 of 3446
Quote:
Originally Posted by zoyd View Post

People often mix up the terms 1D and 3D. A gray scale calibration is called 1D because it corrects only one axis of your color space. The CMS controls on your display (hue, sat, lightness) are actually 3D but they contain only 6 control points. These 6 control points usually aren't enough to remove all errors with respect to rec.709 or whatever color space you are aiming for. The madVR 3DLUT(software) and eeColorbox 3DLUT (hardware) contain many thousands of control points that can be used to correct color. The basic process work flow is:


Characterize your display errors (profile) -> calculate what adjustment need to be made to the source to remove errors at the display->Load those adjustments into your 3DLUT device.

Ooh! So, so CalMAN v5 can measure & correct 3D LUT and apply it to Windows OS or player like madVR, but a hardware box will actually overwrite HDTV LUT? Permanently? That is cool! But considering that this is a 40" 1750:1 CR LCD display from 2009 that cost me $700 back then, I will save up for a good plasma TV before getting even more toys for this hobby!

Thanks again!
post #2644 of 3446
WoooW! I found more options in service menu that deal with mid-colors and some specific to flesh tones! I made one slight change that only slightly moved red saturation in a proper direction and also made orange more accurate and BAM - no more pink faces!

I can only imagine how amazing the picture would be if I could get all saturation and all ColorChecker measures on target! I know CalMAN v5 can measure all these things, but can it actually correct all the colors, hues and saturations accurately and let me save it all to an ICC profile (or any other profile type) that I can use in madVR or in Windows??? I know it will not be a hardware approach, but I got exhausted fixing these things, one tiny change leads to way more changes and it goes back and forth, it can take weeks to get this very accurately if all I use is my service menu...
post #2645 of 3446
Thread Starter 
Quote:
Originally Posted by MonarchX View Post

WoooW! I found more options in service menu that deal with mid-colors and some specific to flesh tones! I made one slight change that only slightly moved red saturation in a proper direction and also made orange more accurate and BAM - no more pink faces!

I can only imagine how amazing the picture would be if I could get all saturation and all ColorChecker measures on target! I know CalMAN v5 can measure all these things, but can it actually correct all the colors, hues and saturations accurately and let me save it all to an ICC profile (or any other profile type) that I can use in madVR or in Windows??? I know it will not be a hardware approach, but I got exhausted fixing these things, one tiny change leads to way more changes and it goes back and forth, it can take weeks to get this very accurately if all I use is my service menu...

I don't know if CalMan supports creation of madVR LUTs but if they do it will be accurate (ask in one of their threads). ArgyllCMS supports it and directions on using a gui front-end to do this are in the thread I linked. Unfortunately windows itself will not use the 3DLUT so it will only be useful for your madVR rendered material.
post #2646 of 3446
Quote:
Originally Posted by zoyd View Post

Quote:
Originally Posted by MonarchX View Post

WoooW! I found more options in service menu that deal with mid-colors and some specific to flesh tones! I made one slight change that only slightly moved red saturation in a proper direction and also made orange more accurate and BAM - no more pink faces!

I can only imagine how amazing the picture would be if I could get all saturation and all ColorChecker measures on target! I know CalMAN v5 can measure all these things, but can it actually correct all the colors, hues and saturations accurately and let me save it all to an ICC profile (or any other profile type) that I can use in madVR or in Windows??? I know it will not be a hardware approach, but I got exhausted fixing these things, one tiny change leads to way more changes and it goes back and forth, it can take weeks to get this very accurately if all I use is my service menu...

I don't know if CalMan supports creation of madVR LUTs but if they do it will be accurate (ask in one of their threads). ArgyllCMS supports it and directions on using a gui front-end to do this are in the thread I linked. Unfortunately windows itself will not use the 3DLUT so it will only be useful for your madVR rendered material.

MadVR 0.86.11 release added import option for eeColor 3dluts (e.g. from LightSpace CMS)

You can export eeColor LUT's by using LightSpace HCL or CalMAN 5 Ultimate Software solutions.
post #2647 of 3446
Quote:
Originally Posted by ConnecTEDDD View Post

MadVR 0.86.11 release added import option for eeColor 3dluts (e.g. from LightSpace CMS)

You can export eeColor LUT's by using LightSpace HCL or CalMAN 5 Ultimate Software solutions.

Yep, exactly. What's still missing is for either LightSpace or Calman to add support for madVR's test pattern generator. Doing that would make things quite a bit more comfortable. The Calman guys already said that were going to look into that sooner or later. But they've been saying that for ages and never did anything yet. Whether LightSpace adds support for madVR's test pattern generator probably depends on how many LightSpace users request it...
post #2648 of 3446
Quote:
Originally Posted by zoyd View Post

Regarding further improvement on the HTPC, you can correct all of your gamut colors to rec.709 (or any other colorspace within your display's gamut) using a 3DLUT.
Fantastic. You think you have a fairly good grasp ov how things work. And Wham! there you go. I had no idea you could improve windows video rendering in a HTPC to that degree. Will have to look into that for my HTPC setup for sure. Thanks for the link Zoyd!
post #2649 of 3446
Quote:
Originally Posted by madshi View Post

Quote:
Originally Posted by ConnecTEDDD View Post

MadVR 0.86.11 release added import option for eeColor 3dluts (e.g. from LightSpace CMS)

You can export eeColor LUT's by using LightSpace HCL or CalMAN 5 Ultimate Software solutions.

Yep, exactly. What's still missing is for either LightSpace or Calman to add support for madVR's test pattern generator. Doing that would make things quite a bit more comfortable. The Calman guys already said that were going to look into that sooner or later. But they've been saying that for ages and never did anything yet. Whether LightSpace adds support for madVR's test pattern generator probably depends on how many LightSpace users request it...

Until LightSpace's Dev. Team can add support for the new madVR patch/pattern generator (if required by some users), next month the users will be able to do this by using MP4 Media Files Edition Of my Disk that I'm preparing now and it will be ready in December.

Currently the disk has been released as a Blu-Ray Disk Project only.

More details about the disk contects can be found here.

This can be done by using LightSpace's DIP Mode with Ted's LightSpace CMS Calibration Disk Chapters that work as a pattern generator for 10/17-Point Cube Profings (1.000/4.913 Color Points)

Here is a recorded video that proofs the perfect Sync between the Ted's LightSpace CMS Calibration Disk as a Source, Played from Media Player Classic - Home Cinema with MadVR as a Renderer for the display of the 4913 Colors of a 17-Point Cube Profiling.



I opened the DIP Mode (7sec) Chapter of my Disk with Media Player Classic HC, Media Player Classic Window, and then I pressed only one Synced Mouse Click and after 9 hours and 22 minutes it's still Synced Perfectly!

Check that picture before playing the video:



To help you about what you are seeing, the BLUE HightlighTED Box is the LightSpace Internal Pattern Generator Window....

The RED HightlighTED Box is the Media Player Classic - Home Cinema with MadVR as a Renderer that is playing the Chapter Video i'm using for the Disk, this window supposed to be maximixed to full screen when you are profiling from Monitor/TV....

The GREEN Arrow indicates the 11.11% Window Pattern supposed to be centered at yout Profiling Display...

*** I used DIP Mode 7 Sec Chapter only for demonstration of the Perfect Sync between the sofware and my Disk. DIP Mode time depends on the probe and the display you are profiling, and might be shorter than 9 hours. To my KURO Display with iD3 using 1.5sec Probe Intergration Time, i'm using the 6 Sec DIP Mode Chapter (8 Hours). This is because KURO has the deepest Black and this is taking more time for the iD3 meter to read it, so to stay in Sync i had to use that Chapter. Dark shades generally is taking more time to be measured...***


*** EDIT *** : This Video was recorded @ January 2013, when I was testing my Disk's LightSpace DIP Mode Chapters, with fixed displaying of each color Point of the total number of 4.913 (17-Point Cube)....synced with LightSpace.

The Disk has 1,2,3,4,5,6,7,8,9,10 seconds per Color Point Chapters for LightSpace. This gives a wide support for Fast/Normal or Slow Meters that can use the Disk as a Source for all these Patterns. for Automated 17-P Cube's without any other user prompt, exept the initial mousy click sync of selecting 'Start Measurenments' from LightSpace Software.
post #2650 of 3446
Thread Starter 
^^This post is more appropriate in your thread, a link to it would have been sufficient.
post #2651 of 3446
Quote:
Originally Posted by zoyd View Post

^^This post is more appropriate in your thread, a link to it would have been sufficient.

My post is a modified version of the original post...if you read the original makes no sence, we are talking now for MP4 files.
post #2652 of 3446
Wow... I just did a color check on CalMAN v5 and there was a lot of discrepancy in green, cyan, and blue. So, I adjusted using CalMAN v5 and it shows Detlae E of 2 at most, but in HCFR its WAY off. So, who do I trust when measuring color space? Its one or the other... I used Recommended Color Checker and also CIE2000, but still, CalMAN v5 has its own recommendations. What about ChromaPure? Is it closer to CalMAN v5 or closer to HCFR? I am sure they use different formulas...

I noticed that when I adjusted with HCFR first time, faces were pink, so I adjusted my reds again and faces stopped being pink but they seemed a bit pale. However, once I adjusted with CalMAN v5, faces became more natural-orange-ish!

What is the general consensus? I can decide on what I prefer, but I want the most accurate picture...
Edited by MonarchX - 11/14/13 at 6:25pm
post #2653 of 3446
MonarchX,

Some quantitative data would be helpful for us to understand what you are observing.

Larry
post #2654 of 3446
Quote:
Originally Posted by MonarchX View Post

Wow... I just did a color check on CalMAN v5 and there was a lot of discrepancy in green, cyan, and blue. So, I adjusted using CalMAN v5 and it shows Detlae E of 2 at most, but in HCFR its WAY off. So, who do I trust when measuring color space? Its one or the other... I used Recommended Color Checker and also CIE2000, but still, CalMAN v5 has its own recommendations. What about ChromaPure? Is it closer to CalMAN v5 or closer to HCFR? I am sure they use different formulas...

There should be no difference in the readings if the same corrections are applied in the different programs apart from minor differences due to differing algorithms, rounding errors and so forth. How do you setup your sensor in these programs?

One pitfall in HCFR is that it is possible to apply a correction matrix in the first sensor dialog on startup, and then a spectral correction in the next Argyll dialog. Only ONE of these should be used at one time. Actually there is a third way by applying a manual matrix. All three are mutually exclusive. Choose ONE at a time or you get double corrections that puts you way off. I did that once by mistake and got confused too.
post #2655 of 3446
I am another one of those Chromapure users with a sensor that has a correction report with it. I always thought that the difference between the ccss and direct speciifc matrix correction would be minor at best with the D3, so I have postponed using that until now when I reached the best calibration I could get with only ccss. I used JohnAd's Excel sheet that transforms the xy readings from my Chromapure report to xyz matrix form and entered it. And yes, the changes WAS minor. Blue showed to high and red to low. The difference in setting was blue -1 and red +1 in both high and low settings. There was some minor changes in CMS as well, but nothing notable.

What these little changes did however was breathtaking. It's a world of difference! The coldness of the image was gone and replaced with that D65 neutral warmth I lacked all along. Color nuances and skin tones fell in place. The linearity errors that this TV has with fluctuating saturations and luminosity over changing APL is still there but less noticable.

I think Tom H doen't want us t share these correction reports so I will not post them here. I would not believe them to be of much use either way since ithey are specific to my personal sensor. It may be that my D3 is a tiny bit different than the generic one, or that the ccss file isn't good enough. I guess a bit of work in this area by someone with a good spectro and access to a number of different displays would be beneficial...
post #2656 of 3446
What tv do you have? I'll guess a Panasonic?
In my st60 changing one click in the cut- off's will have a major impact on the wb. Unfortunately, I have no possibility to correct my i1 against a specto and what I've noticed is that with edr for plasma the readings have to much red over the general plasma corrections that zoyd included into hcfr.
post #2657 of 3446
Thread Starter 
Quote:
Originally Posted by MonarchX View Post


What is the general consensus? I can decide on what I prefer, but I want the most accurate picture...

Both programs have a long heritage and have been cross-checked extensively - when used properly they will yield the same answer. Other than ensuring exactly the same test conditions, probe corrections, etc. there will be some variation due to probe and display stability. This will typically not exceed 0.002-0.003 in x,y and perhaps 0.3-0.4 in dE 2000. One other potential difference although I'm not sure how Calman handles it, is that the color checker and other saturation patterns were originally designed to be used on a display with a flat gamma response of exactly 2.2 So if your display does not have that gamma response there will be some small inaccuracies under this assumption. HCFR handles this by calculating all the color checker and saturation targets based on the user's measured average gamma, this will be more accurate than assuming a 2.2 gamma for displays that differ from that target.
post #2658 of 3446
Please disregard this as I consolidated everything in one post below!
Edited by MonarchX - 11/15/13 at 11:46am
post #2659 of 3446
Again, I apologize for wasting precious forum space, but you should look in the post below for all the information and explanations!
Edited by MonarchX - 11/15/13 at 12:01pm
post #2660 of 3446
I am making too many replies, so I decided to consolidate the issue into a single post.

TV - Samsung LN40B550 CCFL LCD S-PVA Panel from 2009. It has 2 gray scale controls and Saturation + Hue controls for all primary and secondary colors.
Colorimeter - i1Display Pro from X-Rite with latest drivers.
Software - CalMAN v5 with latest updates and driver pack, HCFR 3.1.0.2. I used identical settings for measuring RGB WB and color gamut, focusing mostly on 75% saturation, but attempting to get all of the saturations correct.
My luminance gamma is a straight line 2.2. My red and blue gammas, however, are off, but they should not matter. Both applications - CalMAN and HCFR were set to use pure power curve gamma of 2.2.
Room light was not taken into consideration for any tests.
RGB WB and Gamma are reported the be the same in CalMAN and HCFR. Both are well balanced.
No correction matrices were used in either application as I tested my colorimeter and it was accurate.

CalMAN report. It shows everything is within acceptable range for all colors.



HCFR reports, using the same settings that achieved the above CalMAN results.






CIE2000 - notice that green, blue, and cyan are way off!


Red, Yellow, and Magneta targets and results are identical in both applications, but blue, green, and cyan differ in readings/measures. CalMAN provided Delta E CIE2000 reports for those colors, but it did not show other Deltas, like L, H, and C in those reports..

1. Why is there a difference and which report should I trust to achieve the most accurate picture? Which report would you trust?

2. Why is it that in CalMAN, Delta C (saturation) and Delta H (Hue) can be completely off, yet Delta E will report an acceptable number? Both programs seem to detect the same saturation and hue measures, targets, readings, numbers, but the way they decide on Deltae E must be different. Could it be that CalMAN takes in consideration Delta L (luminosity) ?

3. What would provide a better overall picture - targeting 75% stimulus to have the lowest possible Delta E OR targeting the lowest Delta E for all 25-50-75-100 saturation/stimulus levels, even if it raises Delta E of 75% saturation to a maximum acceptable 2.5-3?. If I target 75% and achieve Delta E below 1, then it may screw up all other saturations, but if I target all 25-100% saturtaions at once/in general, then 75% saturation is going to have a higher (yet acceptable) Delta E, but 25, 50, and 100% saturations will be more accurate.


I am so tired of changing settings. I can't wait for the day when I can just set it using either HCFR method or CalMAN method and just sit to watch some movies!!!
Edited by MonarchX - 11/15/13 at 12:02pm
post #2661 of 3446
Thread Starter 
I don't see any disagreement between the Calman and HCFR CIE diagram, the primaries are in the same locations. The HCFR green location looks a little worse only because the scale is expanded. You don't show any saturation measurements on the Calman diagram so I can't compare those. The dE values you show for Calman for the primaries and secondares, what are the HCFR values? Why haven't you done the same saturation scans in Calman? They should look the same as those in HCFR. You need to understand what you doing before trying to nitpick.

In deciding which settings to use I would use those which minimize the color checker set.
Edited by zoyd - 11/15/13 at 12:22pm
post #2662 of 3446

 

Hi,

 

how can I tell HCFR where to look for new Display Types??? Or where does HCFR expect them to be?

 

I created a new .ccss correction file for my i1 Display Pro and put it everywhere I could imagine (eg. \ProgramData\ArgyllCMS\   and  \%user%\AppDataRoaming\color  and even in the HCFR folder) but HCFR doesn't give me the option to select it as Display Typ for my i1D3...

 

Any ideas??

post #2663 of 3446
MonarchX,

Pretty pictures are not really helpful. I suggested that you supply us quantitative data. Just as zoyd said, I cannot see any major difference in the gamut and the primary points on both diagrams seem to be the same.

Larry
post #2664 of 3446
The HCFR vs CalMAN issue is settled - they both provide exactly the same saturation and hue data.

But at this point all I care about is:

What would provide a better overall picture - have the lowest possible Delta E for 75% saturation (REC. 709 [75%]) OR targeting the lowest Delta E for all 25-50-75-100 saturation levels, even if it raises Delta E for 75% saturation ABOVE acceptable Delta E of 3?. If I target 75% saturation and achieve low Delta E, then it screws up 100% saturations (oversaturated), but if I target all 25-100% saturtaions at once, then 75% saturation is not going to have an acceptable Delta E, but 25, 50, and 100% saturations will be more accurate.

CMS_1.xls 29k .xls file

The above attached file is HCFR Excel data from a calibration attempt with HCFR to get the lowest possible Delta E for REC. 709 (75%) and I mostly succeeded on getting Delta E of 3 or less.

I just ran CalMAN saturation sweeps and they provided exactly the same results as HCFR at 25, 50, 75, and 100%

So, is that the general advice - getting 75% accurate? The result is that my 100% primaries are now oversaturated. I can't get 75% accurately if I want to get 100% accurately.
Edited by MonarchX - 11/15/13 at 5:29pm
post #2665 of 3446
What are the three primary color gamut xy coordinates at 50%, 75%, and 100% amplitude from the results using each software program. This will give us quantitative values of saturation and hue. If I understand you correctly, you are saying that these values are the same then there is no difference -- a pretty obvious conclusion.


edit: my post is now moot.


Larry
Edited by LarryInRI - 11/15/13 at 6:27pm
post #2666 of 3446
Thread Starter 
dE 2000 for colors is calculated the same way in both programs and includes all three components dL, dC, and dH. If you'd like to view the exact formula you can find it at brucelindbloom.com under the math selection.

Typically minimizing average dE over all saturations (including color checker) is preferred over getting any one level perfect.
post #2667 of 3446
Check this out!

This is from "Saturation Sweeps" section of the Quick Analysis workflow in CalMAN v5:


Yet, this is from "Colorspace" section:


These are from "Luminance Sweeps":


ColorChecker:


Just in case you missed, below are my RGB WB, gamma, and gray scale:





Why is Saturation Sweeps dE so different from Colorspace dE?

What do you think of these results? I am quite discouraged... I don't think I can do any better... DDC is not supported for this TV, its an old LCD from 2009.

I do know that the picture now looks a hell of a lot better than with any other colorspace adjustments I have made. Faces finally look realistic and its like everything gained more detail. But I shouldn't kid myself, right? What I am seeing must be nowhere of what I should be seeing.

A buddy of mine told me that it is better than what most pro calibrators would provide for a $400 home visit with exception of providing day & night modes that I have not done yet.
post #2668 of 3446
Have you tried lowering general color control and recalibrate? Also, sometimes using 75% s100% a works better in terms of picture quality than 75-75. Also, using wrong patterns may get you tha same results. For your display you should use fields.
post #2669 of 3446
Thread Starter 
Quote:
Originally Posted by MonarchX View Post


Why is Saturation Sweeps dE so different from Colorspace dE?

What do you think of these results? I am quite discouraged... I don't think I can do any better... DDC is not supported for this TV, its an old LCD from 2009.

I do know that the picture now looks a hell of a lot better than with any other colorspace adjustments I have made. Faces finally look realistic and its like everything gained more detail. But I shouldn't kid myself, right? What I am seeing must be nowhere of what I should be seeing.

A buddy of mine told me that it is better than what most pro calibrators would provide for a $400 home visit with exception of providing day & night modes that I have not done yet.

They don't look that far apart, probably just meter repeatability.

You've probably done all you can given the controls on the display, dE's on average <= 3 inside the gamut is a good result. If you want to push everything down further you'll have to do it with a 3dlut
post #2670 of 3446
My Orange Yellow is way off @ dE 7.

I do have this SKIN control in service menu that adds a lot of orange, replacing pink in gamut, but it screws up the red gamut at any setting. I tried adjusting Color control and also Mid-Color control but they fix little but brake more...

I found that yellow is the problem here. If I increase yellow saturation dE by adjusting hue, it will bring ColorChecker Yellow Orange & Yellow dEs down. But yellow saturation sweeps for 75% and 100% will be at dE 5 instead of 3.

1. If Yellow Saturation @ 75% and 100% is @ dE 3, then ColorChecker Yellow is @ dE 5, and Orange Yellow is @ dE 7.
2. If Yellow Saturation @ 75% and 100% is @ dE 5, then ColorChecker Yellow is @ dE 2, and Orange Yellow is @ dE 5.

Which choice for more accurate picture? 1 or 2? If you average choices' dEs, then choice 1 average is dE 5, and choice 2 average is dE 4. But isn't yellow saturation simply more important than orange/yellow orange?

I only have CalMAN v5 Control Edition, which does not allow me to make 3DLUTs...

Do you have to have 3D LUT hardware to create a 3D LUT with software?
Edited by MonarchX - 11/16/13 at 7:11am
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Display Calibration
AVS › AVS Forum › Display Devices › Display Calibration › HCFR - Open source projector and display calibration software